No longer supported

This model is legacy (released May 2023) and is no longer actively maintained or recommended by Google. Consider using their current flagship model instead.

legacylegacyllm Proprietary

PaLM 2

by Google· Released May 2023· Cutoff Early 2023

PaLM 2 is a large language model from Google, succeeding the original PaLM. It excels at multilingual tasks, reasoning, and coding, and is available in various sizes (Gecko, Otter, Bison, Unicorn). It has been superseded by the Gemini family of models.

Official Site API Docs 📄 Research Paper

Input cost

$0.50 per 1M tokens

Output cost

$1.50 per 1M tokens

Context window

32K tokens

Max output

8192 tokens

Modalities

text

License

proprietary

Capabilities

Text GenerationCode GenerationMultilingual TranslationReasoningChatEmbeddings

Best For

General-purpose text generation, code assistance, and multilingual tasks with a focus on reasoning.

Strengths

Strong multilingual performance across 100+ languages
Advanced reasoning and logic capabilities
Efficient training with compute-optimal scaling
Multiple sizes for different deployment needs

Limitations

Legacy model, no longer actively developed
No native vision or multimodal support
Smaller context window compared to newer models
Limited to text-only inputs

Use Cases

Chatbots and virtual assistants

Code generation and debugging

Language translation and localization

Content summarization and paraphrasing

Question answering and knowledge retrieval

Creative writing and brainstorming

Educational tutoring and explanation

Improvements Over Previous Model

Improved multilingual capabilities with training on 100+ languages
Better reasoning and logic performance compared to PaLM 1
Compute-optimal scaling for efficiency
Multiple model sizes (Gecko to Unicorn) for flexibility
Enhanced code generation and understanding
Improved safety and alignment through RLHF

Back to all models